Our beloved colleague, my mentor and the person who deserves all the positive adjectives of our planet with her constant smile, Olga Torres has left us. We will always remember and love her.

RT @karpathy: +1 for "context engineering" over "prompt engineering". People associate prompts with short task descriptions you'd give an….

RT @rainfilmfest: 💧 Seguim a la Sala Nau del Campus del Poblenou amb el SIDE EVENT organitzat pel Departament de @TraduccioUPF, parlant del….

RT @psgijon: Estàs acabant un grau o en màster en traducció? Llavors, aquest cicle de conferències t’interessa!! 👇👇.Inscripcions aquí: http….

We advocate multidimensional, cost-aware evaluation protocols and highlight research directions that could tip the balance: leaner coordination strategies, selective agent activation, and hybrid pipelines combining single-pass LLMs with targeted agent intervention.

indicating that reasoning layers capture semantic nuance undervalued by surface metrics. Yet these qualitative gains carry steep costs. The sequential agent consumes roughly five times, and the iterative agent fifteen times, the tokens used by NMT or single-pass LLMs.

while both multi-agent workflows trail. Human evaluation reverses part of this narrative: o1-preview produces the most adequate and fluent output in five of six comparisons, and the iterative agent edges ahead once,.

efficiency with total input-plus-output token counts mapped to April 2025 pricing. Automatic scores still favour the mature NMT system, which ranks first in seven of twelve metric-language combinations; o1-preview ties or places second in most remaining cases,.

, on test data drawn from a legal contract and news prose in three English-source pairs: #Spanish, #Catalan and #Turkish. Automatic evaluation is performed with COMET, BLEU, chrF2 and TER; human evaluation is conducted with expert ratings of adequacy and fluency;.

We benchmark five paradigms, Google Translate (strong NMT baseline), GPT-4o (general-purpose LLM), o1-preview (reasoning-enhanced LLM), and two GPT-4o-powered agentic workflows (sequential three-stage and iterative refinement). .

Large language models (LLMs) and multi-agent orchestration are touted as the next leap in machine translation (MT), but their benefits relative to conventional neural MT (NMT) remain unclear. This paper offers an empirical reality check.

Link to the manuscript:

With my colleague @VicentBriva, we compared translation quality of current automatic translation generation methods using two translation agents (one suggested by Andrew Ng et al. and one developed by Vicent Briva-Iglesias, PhD), a reasoning model, a zero-shot LLM and and an NMT.

Join us online today:

RT @estrans: New book: The Social Impact of Automating Translation.By: Esther Monzó-Nebot & Vicenta Tasa-Fuster (eds.)..

Elkington’s (1997) triple bottom line from business ethics to be utilised for technology evaluation. This talk introduces our proposal, some related previous research, and what it might mean to evaluate while giving equal weight to people, planet, and performance.".

"In light of predictions of job displacement in translation (and beyond) due to generative AI, and the urgency of sustainability in world on course for temperature rises above three degrees, some colleagues and I proposed an adaptation of.

Link:

Using ChatGPT o3-mini-high, I've created a simple HTML file to visualize the content of TMX files. When you want quickly check something in a TM and you do not want to lose time opening it in a CAT tool, you can quickly consult this offline tool. Use it, it is free:

RT @pmarca: Deepseek R1 is AI's Sputnik moment.